OpenAI brings GPT-5.5 and Codex to AWS Bedrock for enterprise AI

GPT-5.5 lands on Amazon Bedrock alongside Codex, giving enterprise teams access to OpenAI's frontier models through the AWS governance and compliance stack they already operate.

The move matters less for raw capability than for operational plumbing. According to CIOL, OpenAI made both GPT-5.5 and its AI coding agent Codex generally available through Amazon Bedrock, the managed service AWS offers for building AI applications with native security and billing controls. For enterprise IT teams, that means procurement, security reviews, and compliance workflows can stay within the AWS ecosystem rather than requiring a separate vendor relationship with OpenAI.

Enterprise AI deployments share a structural bottleneck: the distance between a working prototype and a production system that legal, security, and finance will sign off on. Access to capable models has expanded rapidly; the governance layer has not kept pace. Routing through Amazon Bedrock wraps GPT-5.5 workloads in AWS Identity and Access Management, VPC isolation, and audit logging. That is the operational bet OpenAI is making here.

The distribution play

The Codex inclusion sharpens the story. Rather than positioning GPT-5.5 as a raw API, OpenAI is pairing it with an agent purpose-built for software engineering: generating code, reviewing pull requests, debugging failures, and modernizing legacy systems. Development teams already running CI/CD pipelines on AWS can now add AI-assisted capabilities without standing up a separate integration layer.

Pricing data from Price Per Token places GPT-5.5 Short Context at $12.50 per million input tokens and $75.00 per million output tokens via Azure. AWS Bedrock pricing has not been officially confirmed at those levels, but enterprise customers should expect comparable economics plus Bedrock's managed-service margins. At those rates, GPT-5.5 is not a commodity play; it targets high-value, low-volume inference where output quality justifies the cost.

LLM Stats shows GPT-5.5 Instant, a lighter variant in the same family, released in early May 2026. The AWS announcement covers the full GPT-5.5 as the primary offering, with Instant as a lower-latency option for less demanding tasks. AI Release Tracker, which has catalogued 160 frontier models since ChatGPT's November 2022 launch, illustrates just how compressed the current release cadence has become: multiple frontier-class models per month, from multiple labs, all targeting enterprise budgets.

What this means for practitioners

The real story is distribution as competitive strategy in artificial intelligence. Model quality at the frontier has converged enough that deployment ergonomics increasingly determine which tools get adopted at scale. OpenAI's arrangement with AWS follows the same logic that led Anthropic to list Claude Opus 4.8 on both Amazon Bedrock and Google Cloud within the same week, and that has Microsoft channeling OpenAI models through Azure AI. The inference layer is commoditizing; the contest now is over which cloud's control plane becomes the default for enterprise AI.

For engineers and applied scientists, this shifts the evaluation calculus. Instead of benchmarking GPT-5.5 against alternatives in isolation, the practical question is whether the AWS integration cuts enough internal red tape to accelerate deployment timelines. Organizations running compliance-heavy workloads on AWS, including financial services, healthcare, and regulated manufacturing, stand to benefit most, since the alternative means a separate OpenAI API relationship with its own procurement and security review cycle.

There is a subtler issue worth flagging. When artificial intelligence capabilities ship through the same channels as storage buckets and serverless functions, organizational approval processes shorten by design. That clearly suits both OpenAI and AWS. Whether faster approvals represent improved governance or compressed oversight depends entirely on how robust the controls inside the wrapper actually are.

No single partnership announcement will settle the competition for enterprise AI infrastructure. Each frontier model that routes through a hyperscaler's managed service is one more data point that the direct API access era is giving way to something resembling a conventional software procurement cycle. For OpenAI, the open question is whether that makes GPT-5.5 more embedded in enterprise workflows or simply more interchangeable with whatever model occupies the next slot.

FAQ

What is GPT-5.5 on AWS Bedrock?
GPT-5.5 is now available through Amazon Bedrock, letting enterprise teams access OpenAI's frontier model inside their existing AWS environment with native security and compliance controls, rather than through a separate direct API contract.

How much does GPT-5.5 cost?
Comparable pricing via Azure is $12.50 per million input tokens and $75.00 per million output tokens for the Short Context variant. AWS Bedrock pricing has not been officially confirmed at time of publication.

What does OpenAI Codex do on Amazon Bedrock?
Codex is OpenAI's AI coding agent. On Bedrock, it helps development teams generate, review, debug, and modernize code within existing AWS workflows, removing the need for a separate integration layer.

Is GPT-5.5 Instant the same model as GPT-5.5?
No. GPT-5.5 Instant is a lighter, faster variant released in early May 2026, suited for lower-latency or higher-volume tasks. The full GPT-5.5 targets higher-quality inference at a higher per-token cost.